Four Econometric Models and MonetaryPolicy: The Longer-Run ViewKEITH M. CARLSON and SCOTT E. HEIN

NE key element in the making of an infhrniedeconomic policy decision is the accuracy with whichpolicymakers can gauge the longer-run consequencesof their policy actions amid strategies. Crucial to suchattempts to grasp these policy consequences is the useof econometric models. Whether current eeouomnetricmodels are useful in this respect depends upon their“long-run” characteristics; unfortnnately, until recent-ly, there had been virtually no study of the comnpara-tive long-run properties of the major econometricmodels currently in use. Most analyses instead havedealt with how well these models forecast a few quar-ters ahead.

This situation changed with the publication of a re-cent study by the Joint Economnic Committee (JEC) ofCongress that focused explicitly on the economic im-pact of alternative long-run monetary strategies usingthree well-known econometric models. Missing fromthe JEC study, however, was an econometric assess-nnent using an explicit mnomietarist model. The purposeof this paper is to extend the JEC study by comparingtheir results with those obtained ibr the St. Louismodel. Amialysis of the St. Louis model according tocriteria used in the JEC sttmdv is infbrmative for tworeasons. First, it imidieates whether a mnomietaristframework provides additiomial insight into the long—run effects of monetary policy. Second, it providespohicymnakers the opportumiity to comnpare the lomig—runproperties of a nionetarist mnodel with those of themajor mionmonetarist niodels.

F.I~flTUR.E~‘O.F1.’HE J.EC STfl~JDY

The JEC study examnined the simulated perfor—niance of certain key macroecomiomic variables underfour dhiflerent long—rtmn momietary strategies. Threelarge—scale econometric models were amiahvzed: thoseof Chase Eeonomnetries. I)ata Resources Incorporated

Robert F. W’eioH rat mh, T/m red’ Lu,‘geS dO/l’ ,Vor/e/ Sotmo /060 us’ of

Four Money Growth Scermorios. a staffstudy prepared! for the use ofthe So hewn mittd,e ni N-I om,etarv a:mdl Fiscal Fol icy (If the Jo: mitEconom:mid’ Gum:mmmi ittee of Congress (Coverm:mmme:It Prim: tim: g Ofilee,1982).

(DRI) and Wharton. the best-known and mnost widelyused models.

Four separate monetary strategies were consideredover a 10-year simulation period (1982 through 1991),using the fourth quarter of 1981 and an Ml growth rateof 5.8 percent as points of departure:

(1) a sudden deceleration ofMl growth tozero percentin one year, and then held at zero;

(2) gradual deceleration of Ml growth to zero pereemitover a five-year period, and then held at zero;

(3) sudden deceleration of Ml growth to 3 percent inone year, and then held at 3;

(4) gradual acceleration of Ml growth to 10 pereemstover a five-year period, and then held at 10.

In addition, each model’s proprietor was asked to run abaseline projection with freedomn to choose the mone-tary strategy.2

2The haselim:e simmmlations thus m’epresented each models assumnp~tion about the futm:re course of momietary policy as of’ N-larch 1982.These ass mm mnp tmons were as fbllows:

C~li:m:eDItI\Vb:’srlom:

Ml Growth Bate:1982’411

6.3%1.55.2

Tb dl model propm’ietors wd-’re flirther it:st rm meted to sin: mm late cad’l:of the losir monetary strategidls tsviee: first, wihioum mnakimig anyjodgmnemstal ~dIJ1st lIen ts. and! sd,eomi d, making an~a~i~I:st men tsdeemed neeessarv to ensure eon sistenev andl gem: crate res smlts thatwere considered sensible, These adjustmmmem:ts were at time dhsere’tint: of the im:dividual mimodel proprietor amsd illvolved mmo contactwith the JEC stall’. The JEC babeledi these two sets of sim:sulations

p:mm’~’ and moanaged!.

The JEC study coneludled. omm the basis of the pure sinmulatiomms,that mm dim:e ofthe modlels cam m he m: setl in’ themOSdIlyes to (led:idld’ am: mOmmgthe mmi om I et ar’ strategies. (he results oi these pure si m:i mmlat id: mis weretermneml puzzh ml g. ‘‘ because the links I :etwed’n the muon dlv g:’owt hmmd the key mnaeroedl000 mimic variables m’ami d:ommmi te r to histo ‘mealexperiemi ed.

The J ICC comIc1t:sions about the manag~~~ism mmlatio mis wem’e mm:ore

positive. \VhildI thd’re still md’ mi maindld somd’ ineomi sis teneie 5 Withhistorical relationships, the :nanaged simi mulatiomi s were judged toprovidld, a better basis for eomms ide rimmg time lom:ger—rum: policy imm:’

plieatiomms of alter:mative momietary aggregate growth strategies.Tbus. im m the disco ssidlmm to fbI low, omllv the mnaoagedl sill:mmlatio::rd’s ‘Its from th t, large scale models an-’ eonsirlerdld,

13

FEDERAL RESERVE BANK OF ST. LOUIS JANUARY 1983

The St. Louis Model

Th basi tm tu o th t Lou mod I de Th prcee natioi r It tb a ofchang ofxelop d n th 1 1960s h r mnain d e s uti 11 pnee to curr nt and la~gd vain o dema d preunchanged in’ then Th model ‘on i ts off ur cur nt and ha d válu of chang in theq aion ad two mdcii tie ( app nd Th r Ituv pnc of n rg) and am astir ofantu i

I undatuon fthe mod land th b i for mts mon p t dpruc hang D ma d pr ssur m d fined astan t label s the C P eqn ho The growth rat of th owtho output relatu e to the “rowth of highG Pi pecuified s alan lion ofcm r nt and lag ed emplo m nt output Anticipated p ‘ change s a~ahuesofM1growtha thea r tandh gg dsalu wemghteds rnofpa tprice han wthth weightofthe growth ofhigh mpioym nt f deral p nda obtain d by estimating the comp a e Aaa iat atare Th m a an t 1 b h tern pnmarih from fimncti n ofp t inflation On pi t growth real GNPthe estimat d coeffl tents, the s m of th co f- i d t rrnmn d r idnally s a the CNP md ntitficuent on rnone growth about units and be sum ommnal C P growth is the urn of real GNP grosm thofth o fflcm nt oi high emplo) n nt xp nditure and th rate f change of pricegrowth is abo t o

1L onail And r d Ice th N-f I on A Mon ‘tan t The mod I remarnrn quations provmd es mMod IforE 0110 mm. Stah:h ama n th: B ‘ ApnI 1970)pp 25 M: ordi s vlsi hha h mimnad mid d ~i a mat ofthr eoth mm eroeeonomm anmabl s Un-spent a mo in rat -of aug kim ft m th o gn I first mpioym nt us timat d a a fim ‘tion of th gap

dmffe n torso 2) dd’tmo mm g pri . ~ ~ h tween a tual output and high employment out-iou v ale andl)a han em tm a ormproc dur fomordin ry IC i q es to mu am dl sq a f r tim put Th Aaabondrat u afuu tuonofpastinflatuonqluations mmm wh h e al ‘orrehmti i ride t. S K mth ‘d The 4 month eomnmercual paper rat ms a f:netmon

Carlsonaimd Sc tt E H : A Anal i of Modmfl d H Lou: of contempor neous Ml growth nd current andMod I apap prep edfo tIm Spm Comif re ceo emparmgth Pred tm’s P rfor aim eof’Ma con mc Mod I ha ed salu s of hang s in the r lab p ice ofat Wash mmgton Liii’s rsstv a S Lot i April 20 1982). n mgs output and price

Though the imistructomis is crc spc cihed imi termns of espectis l’s s lie: c’as the JEC peeifed \ll targets of‘~I1 mione of the mode Is permitted dli cet comutrol ofthis zero percent 3 percent and 10 pci cc mit ~momietar’s a gregate. Both Chase and DRI specif’s thecomitrol of money gross tli through nomihom : o’ss cal me Simnulation of the St Louis :nodc I fom the Ion r_rumisen es. Thtm nomihorro\vedl re sen es S’s em e m’mnipu mnommetari stm ate nit s outlined imi the JEC studs rclated to aehie’s e the the sircdh \l 1 gron th. F or time quired a smmnmptions shout other exogenous lam ihles.\Vhartoi model the tam get s ariahle is \l2 imistead of potential output as assumed to gross 2.5 pereemit perMl. The \\ harton simulations is ‘me conducted usin sear high-c mplos mnemmt expt n hittmm es to mereist at aM2 target rites of4 peree mit 7 percent amid! ii p ret mit stcads S ~ereent rate and tli ‘ change in the a lath e

prmce of cuergi to he zt’ro. T eh te m mimic a baseline’st:’ate gi an as r, ge’ of the haselimie stm ate ic’s fom thelmmg scale inodds X’s as eomistructe ci. W hat this

DEl Ins iii mtt r,stms e prod dome tlm,mt, Ibis d th :0 td hit \l It Sr mIs amounte el to is as ‘u gradual r e ueti mm: in \ I gross th(i i_tb’s ~ ~ mficdl h~tim( J C ( Ima i_. 0~ith olim( r lmammd. mms~’mltoil , md d rm or proc ( dmmr am d is ss o:m,mhld’ to ad hmmst NI t tar’ ti lion: 5.8 pe rec nt ate in fourth qua: te: lYSi to 5.0pn c.msds percent iim 1991.

‘,mn~’ th smmm:l itmomis s~mc rum: mm \l,mrch 1982 (h tsi_ [‘i_dfli—

ommmd’tmmr.s has nimsiel tlmcmm mind I to ::meor’) mratd mu ‘ :::nmmct Sri

etor to mc fit U cbamm 4. mm: Fedm.r ml list rid polo.\ pro ‘m.climmcs :mm Ml ms Sm ‘nit: ‘c nOmmi S :rmahld mmm tlmc omodhel I o”e’c m. di thit m 5 i

Oe tol Cr I9m I St time tim m tim mmml :ml,mtmon is m mm rim:: I iii ( I m. e iii m s for m.’omm.pam’:n tIm I 55 I, irtomi Inodld I vs m th t lid otlm em mm gic Is I bitnmorIm I 0 edi Iii 0 IdId S )f em dm1 r mtmommmmm a tim pr oman h mm: mm. I o r mmlt, mm: M I s,ro’stim : mk is d m d S ii r.mI Is I) tm I mint prc CisC Ii d 0mm

mndmrmc tar’s mmmlii Oct m:tc ‘mt s mth th JF( s mmm tiutmm mis

14

FEDERAL RESERVE BANK OF ST. LOUIS JANUARY1983

PROPERTIES OF THE MODELS ASREVEALED BY THE SIMULATIONRESULTS

This study follows the general format of the JECstudy, using the U.S. economic experience from 1956through 1981 as a guide in comparing the models. Ifcertain systematic relationships among key variableshave held overthepast 26 years, the simulation resultsfor the next 10 years should be roughly consistent withthat experience if one is to place much faith in themodel. Deviations from historical experience place theburden of explanation on the individual model pro-prietor.

Simulation results relating money growth to (1)nominal GNP growth, (2) Inflation and (3) real outputgrowth are considered first. Then, the relationshipsbetween real output growth and unemployment, andbetween nominal interest rates and inflation are evalu-ated. Since the longer-run relationships are ofprimaryinterest and since short-run adjustments make the re-sults difficult to interpret, the results lbr the last fiveyears of the simulations, 1987—91, are investigated.

GNP, Money and Velocity

With simulations of the fbur long-run monetarystrategies and a baseline simulation, five observationscharacterizing the 1987—91 period were generated lbreach model, providing a basis fbr examining the rela-tionship between money growth and nominal GM’implicit in each. This relationship is rekrred to con-ventionally as the velocity of money. The well-knownequation of exchange portrays this as

MV Y, or V

where M is money stock, Y is nominal GNP, and V isthe velocity of money. In its growth rate fbrm,

Although velocity growth is influenced by manyvariables, it has shown considerable stability duringthe 1956-81 period. The implication ofthis stability isthat, in the long run, nominal GNP growth is relatedclosely to the growth of Ml. The stability of velocitygrowth further suggests that a 1 percent change in rateofgrowth ofmoney should coincide generally with a 1percent change in the rate ofgrowth of nominal GNP.

The large-scale econometric models do not speclfrGNP as a direct function of money. In these models,

money affects GNP indirectly via interest rates andwealth or real balance elfrcts. Despite this, the largemodels still yield systematic relationships betweenmoney and GNP.

Chart 1 summarizes the money-GNP simulation re-sults. Each model is summarized by plotting the aver-agegrowth ofsimulated nominal CNP fur the 1987—91period against the average growth rate of Ml lbr thesame period. Eachpoint represents model results for aparticular long-run monetary strategy.5 As notedabove, these strategies are stated in terms of Mlgrowth, and include (1) a sudden deceleration to zeropercent, (2)agradual deceleration to zero percent, (3) asudden deceleration to 3 percent, (4) a gradual accel-eration to 10 percent, and (5) abaselinestrategychosenby the model proprietor.

The historical line is derived by regressing the five-year average growth rate of nominal GNP on the five-year average growth rate of Ml. The parallel linesdepict the regression estimate plus or minus one stan-dard error of the equation. Ifvelocity growth is totallyindependent of money growth, then the slope of thehistorical line would be 45 degrees. The estimatedslope, in fact, is not significantly different from 45degrees.

Comparing the different models with historical ex-periencesuggests thatnone ofthe large-scale models isgenerally consistent with the actual past. Only Ibur ofthe 15 simulated cases for these models fall within thehistorical band. The DIII and Chase simulations indi-catethatvelocity growth is related negatively to moneygrowth, so that higher rates of money growth do notyield proportionally higher nominal GNP growth. Onthe other hand, simulation results for the Whartonmodel indicate that higher money growth results inmore than a proportional increase in GNP growth. Thisresult, however, Ibliows from the nature of the finan-cial sector in theWharton model. On the basis of M2,which is Wharton’s actual monetary target variable,velocity growth is related negatively to money growthas in the Chase and DRI models.

Not surprisingly, the St. Louis model falls clearlywithin the historical band; after all, the GNP equation

5Forthe Ml growth rate associatedwith each strategy, refer to theaccompanying table. Thepointson thechartareconnected foreachmodelin ascendingorderofMl growth. Consequently, the resultsfortheChaseand Wharton models arenot charted with theJEC’sslowest growth strategy farthest to the left. See also footnote 3.

15

FEDERAL RESERVE BANK OF ST. LOUIS JANUARY 1983

V

Table IGM’ Money and Vetoetty (1981—91)

Average Ann at Reaufta

Mode and Sirategy M V

Chase1 8% 1k2 1 80 69

a as S4 106 11 10Baseline 64 94 30

urnI CD 7 7,22 0,0 73 7

30 914 100 15 3Baseline 4 96 5,3

1 30 SE 282 15 64 4~B3 3,2 75 424 65 15 5Baseline 49 aS 46

St Louis1 00 28 282 01 20 28

3 624 99 14,0 37Baseline 52 8 3

is constructed to he consistent with this historical inflation is related directly to mone~giowth. In t rmsexperience.t’ The proprietors of the other models ofl~’r of the equation of xchange with rates of change ofno explanation as to why their models predict that prices and output ( P X ) suhstituted fbr Yvelocity behavior in the future will he different fromthe past.

A justification of the money—inflation relationship isInflation and Mone~1 that V and X are not related systematically to M over

the long run. Consequently, variations in M eventuallyEconomists generally agree that, over the long run, are reflected in F

Fo evaluate the money—inflation relationship for the

tse St. Louis model simulations tic, show a weak positive rela- different models, the simulation results are suinma-tsonslsip between velocity growth alIt

1morse) gros.vth. ‘ibis result

Occi rs hecause the estimate’1 sous of the eoeflmcieuts on M inthe ONP equatio ii s slightlv greater than uiii tv.

TThe JEC study suggests that the reason the large—scale models ron

contrarv to hsstorieal velocity experience is that they are his It toshort—rum’ specifications, that is, their lbcs is is on fbreeasting forshort periot’s is ito Hi e hat tire. Such an explanation ssiigls t hessppi’opi’iate Fr the Chase model, bcit the I)R1 anti Wlsartsss i imsdelsare annual models. The resoIts s tigges t that something more furs —

dament~slis awry. irs addition, the St. L,o tsis model which is aquarter

1v inodel, does s sot exisihit any departure fI’om Is i storical

Ion g—ri in velocity behavior.

± = +

For example, Barro arid Fischer introduced their 1 976 ssirvev ofmonetary theory with the fkAlowing state neil t;

‘‘Perhaps tlse issost st rikissg essrsts’as t between current views ofusonev and those of 30years ago is tlse rediscovery of the endostencityoF the priec level and mu ation usd their relation to the behavior ofmoney.

Robert j . flarro and Stars 1ev F’iselser, ‘‘Recess t I )evelopsssents inNionetarv Theory. ‘ Jous’natof Marsetaoj Eeononhlc.s tApril 1976),p. t33.

chart i

Money and GNP

‘V(Percentl14

13

12

II

10

0 1 2 3 4 5 6 7 8 9 10M

(Percent)

16

FEDERAL RESERVE BANK OF ST. LOUIS JANUARY 1983

Chart 2

Money and Inflation Table 2

Inflation and Money (1987—91)Average Annual

Resuits Emna Year

Model and Stratecy M P F

chase1 30’c 48°c 4902 49 443 28 52 484 ~06 78 79Baseline 4 6.3 5 9

DRI00 40 36

2 00 40 353 30 61 581 10.0 104 98Base.nc 41 65 62

Wharton1 30 29 212 15 34 733 32 42 364 65 98 103Baseirne 49 66 62

St. Louis1 00 27 17

2 01 09 193 30 16 284 99 ‘00 128Baseline 52 52 60

rized in chart 2. Without exception, all four models a sensitivity that appears too high. While the modelsshow a direct relationship hetween monetary growth generally are inside the historical band for moneyand inflation. There is substantial variation, however, growth rates in the neighborhood of the 1956—Si aver-in the degree of sensitivity among the models. The age of 4.7 percent, a wide range of results occurs forChase model shows a difference its inflation forecasts of monetary strategies that lie at the extremes of his—only 3.5 percent between the slowest and fastest torical experience.9

monetary growth strategies. DRI shows a 6.2 percent -_______

age point diflerential and Wharton a differential of 8.2percentage points. The St. Louis model shows thelargest differential of 14.5.

To provide a basis for historical comparison, theinflation rate was regressed on the average of moneygrowth over the previous five years for the 1956—Siperiod. Comparing the simulation results of the fourmodels with this historical line suggests that there issome bias ineach. The Chase and DRI models exhibit asensitivity of inflation to money growth that appearstoo low, while the Wharton and St. Louis models show

°Anexplanation of these diverse results would require a detailedanalysis ofthe inner workings ofeach model. For the most pai’t, thelarge—scale models estimate the price level prinsanily’ by marking upsome measure of labor costs. Consequently, the insensitivity ofinflation to money growth developnsents in the Chase and ORImodels might he related to the stickiness of wages. Tins explana-tion does not seem to explain the Wharton results, however, TIseWharton model shows eonsiderahle sensitivity in the 3 percent to 7percent range for money growth, yet the price determinationprocess apparently is similar to that for Chase and DRI. The St.Lomus model differs from the large—scale nsodels in that prices aredeternsined directly hy demand pressure and past prices. Theinfluence of past prices tends to capture efihets operating throughwages, vet inflation remains sensitive to money growth throughoutthe full ntnge.

o9i’ Hjsioricai reiaiionshmp:

1.46,1,3614 A’-’o.6212.05) t9.70b SE 1.22

-3 ..s. I I

0 1 2 3 4 5 6 7 8 9 10 iiM

(Percent)

17

FEDERAL RESERVE BANK OF ST. LOUIS JANUARY 1983

Chart 3

Money and Real GNP

M

Table 3

Real GNP Growth and Money (1 987—91)A~eracjeArm1LaPesjlla

Mccc. a’lo Slra:eqv M X

Chase3I~~

2 303 28 294 106 34BaSeline 64 ~ U

DRI1 00 302 03 323 30 294 100 28Baseinc 41 29

Wharton1 30 282 15 293 32 324 65 25Basei:ne 49 30

St. Louis00

2 01 392 30 454 99 37Basenr’e 5 2 3 3

Real avp and Mon.ei, terms. Given more time to adjust. the St. Louis modeltends to approach about 3 percent real growth. regard-

A corollary to the long-run, money-inflation rela- less of money growth. - -

tionship is the hypothesis that the trend growth of real -

GNP is not systematically related to long-term moneygrowth. Money may affect the growth of real GNP inthe short run, but if inflation rises one-for-one withaccelerated money growth, as the equation of ex-change indicates, there are no cumulative efl~ctsonreal GNP.

Chart 3 summarizes the money-real GNP rela-tionship from the simulations of the four models. Thethree large-scale models all show real GNP growthrates in the neighborhood of 3 percent, regardless ofwhich monetary strategy is considered. The St. Louismodel, on the other hand, shows greater variation ofreal GNP growth among the strategies. This is becausethe dynamic lag structure of the St. Louis model is suchthat, after 10 years, the model is still a considerabletime away from steady-state equilibrium in growth

The historical line in chart 3 is based on five-yeargrowth rates ofboth money and real GNP. The slope ofthe line is not significantly different from zero, and thestandard error is quite large relative to the mean. Theresults for the three large-scale models are virtctallyidentical. Relative to the large-scale models, the St.Louis model is the outlier, though four of the fivesimulated observations are well within the historicalband; only the strategy ofsudden deceleration of M itozero yields real output growth that is outside the his-torical band. Again, this makes sense because of thelong adjustment process in the St. Louis model; veryweak output growth in the early years under the zeromoney growth strategy is offset by very strong outputgrowth in the 1987—91 period.

In general, the simulation results suggest thatmoney has a neutral effect on real output growth in the

ii and X are average annuam raies for 1987—ti.Historical relationship:

— 3.87 0.07M R - — 0.02

tl.42) (0.72) SE — 0.89-~ I I

0 I 2 3 4 5 6 7 8 9 10

(Perc,et)

18

FEDERAL RESERVE BANK OF ST. LOUIS JANUARY 1983

Chart 4

Real GNP and Unemployment RateAU(Percent)3

AU(Percent)

3

0

—I

-2

-3

.4

-s

-6

-1

long run- A sustained change in the money growth ratehas little or no efl~cton the long-run growth rate of realGN P.

Real GNP a-ad L’ne-mploy-me-nt

Another relationship of interest in macroeconomicsis the one between real GNP growth and the unem-ployment rate. All three of the large-scale models showessentially the same rates of real growth for each of themonetary strategies. Thus, Okun’s law, which relatesunemployment to deviations of actual from potentialoutput, suggests that the change in the unemploymentrate would be approximately equal fir all strategies. ~

Such is not the case. Each of the large-scale modelsshows considerable variation in the change in the un-

‘°ArthurM, Okun, “Potential CNP: Its Measurement and Signifi-cance,” 1962 Proceedings ofthe Bvsines,s and Economic StatisticsSection of tile American Statistical Association, pp. 98—504.

Table 4

Real GNP Growth and Unemployment(1987—91)

.A:erage Aniunl Rato Chaicer. U

2 Mrae: anu Srratc~y X 1985 9

Chase1 s-u~ cot-2 33 05

29 C’4 34 23Buscl:ne 3 0 I 3

DRI30 15

2 323 29 094 26 00Baseine 29 06

Wharton- 28 19

2 29 43 32 324 25 14Baselnc 3 0 0

St. Louis1 55 662 39 2.93 43 454 37 30Baseli.nc 3 3 2 0

i.Illii\ riiiiil jilt iltsjiit lli.li_IIIii.ll ,ilis iiliiai (“I’— I iit I’ ni l~ ii iii c-i,iil ‘ liii ~ I neil

quii~~tlir.itisliiiliii l;Ilun foie..iiiiii,llii.iiIiliiiiiiiiarils

lHiienhi.il cinlj~ut In hew nudeR. \i-’ieiUivlyssInn tin 5~rat .~iu.~ I lull Slit il , U ii r~,nh fileR lu

ii’.uit’ i’i.i slicicli-il ciii-ihi’;ilii,ii i,f I ~ii~~~lIi Ii) /l-il)

i—ti tuc Ii iii ii’ i i-I i,i ii — ii liii iiii(’ifi~ i’~lilt iii il t liiihit I.’J1i1.iu.nt.tii.piitdt’eh.hii

tilt \\ ii:ti’ti)ii nil}dVl. liii -e,nit’, iii- the n1i1nisiti’ e\—if I Iii 1() l)t’ii’ilt. ‘lii’s’s

‘‘sill ttii;L(ei \aii:limi fiiuii i 2:; ~ l11j11e (hiptill iiilviIi()li~ iiilit iii fi)i’ ti ( li.L’( nubIle1

nt I. Iiii ii’ ,tsi — ill’ lii’- it—li iii Ililiclil

The St. Louis model also shows considerable varia-tion in the change in unemployment across monetarystrategies; however, this is due to substantial variationin the growth rate of real GNP. All the unemploymentchanges are negative, because the simulated realgrowth rates exceed the assumed growth rate of 2.5

percent for potential GNP. Moreover, because the St.

0 2 3x

(Percent)

19

FEDERAL RESERVE BANKOF ST. LOUIS JANUARY 1983

Chart 5

Inflation and Long-Term Interest RateRL

Table 5

Inflation and Interest Rates (1 987—91)

Model Average AnnJa Resufrs Final Yearand Strategy ~l P RL RS RL

chase1 30°~ 48°i 1’ 5,s 153’. 1070.2 11 49 115 ‘58 1163 28 52 95 106 874 106 78 116 90 127Baseline 64 63 100 84 96

DRI00 40 101 80 95

2 00 40 101 78 953 30 6.1 111 95 ‘074 100 104 145 124 )4 i

Base,ir’e 41 55 ‘14 100 109

Wharton1 30 29 8.5 65 692 15 34 61 62 723 32 42 107 86 924 65 98 160 138 155Bacejine 49 56 123 94 117

St. Louisi no 2? 56 19 462 01 09 1.5 28 583 30 1 84 -19 854 99 100 140 113 161Baseline 52 5.2 i’d 73 115

Louis nsodel simulates very strong 1987—91 real output hdiai-ion and Inteie~iRnte.~growth in conjunction with the sudden deceleration ofmoney growth to zero, sizable reductions in unem-ployment go hand in hand with such a policy.

The historical line in chart 4 is estimated h’s’ regress-ing the change in the unemployment rate over five-year periods on the five-year growth rate of real GNP.The historical band encompasses only one observationfrom the 20 that are charted. The models’ failure toreplicate history may not be as had as appears in thechart, however. Potential output supposedly grewfaster in the 1956—Si period that! it is assunsed to begro’sving in 1987—91. The simulation results suggest animplied growth rate of potential output of 2.5 percentto 3.0 percent for 1987—91, instead of the 3.6 percentrate calculated fbr 1956—81. Nevertheless, the large-scale models show the inverse relationship betweenreal growth and unemployment suggested by Okun’slaw. In contrast to the St. Louis model, however, thedegree of sensitivity is not well defined.

The relationship between inflation and nominal in-terest rates is the final relationship considered. ‘riscinflationary experience of the last 15 years provides anample basis for examining the nature of tins rela-tionship.

Monetary theory suggests that nominal interestrates reflect inflationary expectations. These expecta-tions can be modeled as a ftmction of past inflationaryexperience. The question examined here is whetherthe econometric models incorporate such a rela-tionship.

Chart 5 summarizes graphically the simulation re-sults for inflation and long—term interest rates. ‘[heChase model does riot appear to show any consistentrelationship between inflation arid long—term interestrates. i’he Wharton model displays a peculiar kink atrelatively low rates of inflation, while the DRI and St.Louis models display a strong positive relationship.

ft(Percent)17

16

15

14

13

12

II

l0

9

P Ia average annual raie tar 1987—al; Hf,corporate bond rafe tar 1991.Hiatorltai relaiionship;

IlL -~2.68 -‘- 1.03 ~

15.28) t1 0.04)I I I

0 2 3 4

P(Percent)

5 6 7 8 9 10 11

20

FEDERAL RESERVE BANK OF ST. LOUIS JANUARY 1983

Chari 6

Inflation and Short-Term Interest RateChart?

Misery Index

What is most obvious from the chart is the incon-sistency with historical experience. The slopes of thesimulation results are roughly consistent, but thegeneral level is vastly different. For the St. Louismock-I, the inconsistency arises because of the usc ofthe serial con-elation adjustment in the simulations.\-Vith long—term rates in late 1981 well above the infla-

tion rate, this difl~rentialonly graduall~’disappearsduring the simulation period. It appears that the large—scale models are ibliowing a similar procedure. in thisregard, it seems that most of the models would domuch better at predicting the c/mange ill long—termrates rather than the level itself.

Chart 6 plots the simmulation results for inflation andshort—term interest rates. Again, with the exception ofthe Chase model, the models demonstrate substantialsimilarities. The St. Louis model tends to simulate thelowest level of short—term rates for a given rate ofinflation. Time historical line, as in time case of long—termrates, is below all the model result.s, but the discrepan-

cy is not as great as that for long—term rates. All themodels, with the exception of the Chase model, in-corporate an inflation premium into short—term rates,suggesting that the lower the inflation rate, the lowershort—terns interest rates will be.

THE POLICY IMPLICATIONS OF

THESE SIMUIiA’I’ION RESULTS

The discussion above emphasized the long—rummproperties of econometric models as reyealecl by thesimulation results. What remains to be determined arethe implications ofthese results for long—run mnonetary

policy. From this longer—run perspective, do the mod—els’ simulation results favor a strategy of slow Mlgrowth, fast Ml growth or something in between?

To aid in this assessment. a crude index, called a“misery index,” is constructed to summarize the re-sults. The index is simply the sum-n of the inflation rate

4

P(Percent)

0 1 2 3 4 5 6

(Percent)M

‘~11 8 9 10 11

21

FEDERAL RESERVE BANK OF ST LOUIS JANUARY 1983

Table 6Misery Index (1987—91) -

Av6rageAnnual ResuLt AnaL Year Misery Index

Modelarmd8trategy M P U P U

Chase 0

3.0% 4.0° 0.5% 1L4~2 11 44 98 1423 28 48 88 1364 106 79 2S 102Baseline 64 68 50 0

OR!1 00 36 05 10.1

2 00 36 64 1093 30 58 63 ii4 100 98 66 184BaseLine 41 82 66 127

Wharton1 30 1 76 92 5 2 76 983 32 3.8 55 934 6 0.3 9.1 94Baseltne 49 62 61 12.3

St LouisI 00 1 38 222 01 IS 82 43

30 .6 39 624 9 128 7 146Baseline 52 60 49 109

and the unemplo’s md nt rate ~mtsome point in time. mm anal’s sis of the adjustment path of inflation and umiemnConstruction of such an mndex is of cour e simisplistic plo’s mnemt.~et it pro’s ides ~ neral informnation for em ‘iluating thd’ ihe general Id’s d Is of the misc r’s index for the fhureflect of the alternati’s mnonetar’s strategies. models indicate suhstantial mari-ttion in thd pm edlmcted

— efleets of alternati’se mond tar’s stratc gics. For the slox’sChart 7 summarize s this misc r\ mude for the 198i Mi ~ro’smth scemiam ios thd St. Lotus model is h’s far the

91 period fbr the four econometric models. In °eueral most optimnistic and the Chasd model ms the mostthe simtmlatiou results mndieate that there is a long—run pessimistic. For the fast \11 gi o’s’s-th stm’iteg’s Ch’mse mspa~oflfrom follo’s’s ing a slo’s’s \I 1 gro’s’sth strateg~’the most optimistic -mud W~hartouis nsost pc ssimnistie.results from the Chase model pm ox idc the oni’s exc’ep— Fhus using this set of m-esults a policy makem is con—tion. There se ~nss to be little basis for cimoosmng be- fronted wmth a dmstmm bing dixersit’s of opinion. Yetweemi sudden and gradual thu Icr ition to zero mont’s three of the four models show a defimte pa’s off fromgrom’s th howe’scr because the nsmsers index dmffcrs fbllowimig a strateg~of sIo~to moderate gm o’s’s th of Ml.little when these strategies are conspar d. An c’s aluation of these strategies would in’s oh e a niore detailed -

Ihm simple index or gmnated ~smththe late Arthur Okun ‘dthough This article extending recent work by Robert Wein-hc called it a “di comfort mdc ‘The term “ml er’s index is used traub at the Joint Lconommc Comnmnittee has comnparI j L St in ‘sIon (a at Keyn ian and “sew C/a4sical - ‘ -

Fcononjs (Nc~York ~nm’ r mt’s Puss 1982. ~ 159 simmmlatmon results fiom varmous econometrmc models to

22

FEDERAL RESERVE BANK CF ST. LOUIS JANUARY 1983

thehistorical record ofthe last 26 years. The emphasisis on the longer-run economic impact of alternativemoney growth scenarios. No single model was Ibund tobe consistent with the historical record on all counts.

The simulation results generally show, however, thepositive consequences of following a slow Mi growthstrategy. Higher rates of money growth are associatedwith higher rates of spending growth, which eventual-ly are reflected in higher inflation rates. Using a simplesocial loss Ginction called the misery index, three ofthefour models indicate that, over the long run, unem-

ployment gains, if any, are insufficient to othet the

increase in inflation.

Consequently, this article — like the JEC studybefbre it — concludes that there are no long-run eco-nomic gains from higher rates ofmoney growth. This istrue even though the models run counter to historicalexperience in some important aspects. Moreover, theresults indicate that higher inflation rates are associ-ated with higher levels of both short- and long-terminterest rates, so that interest rates tend to be higherwhen the faster monetary strategies are followed,

23

AppendixRevised Form of St. Louis Model

Iii(’ \ri¼Iulm iii liii SI. Iummi~ flhi)dVl ii’,i’d (ii lw Li1II~(lflin isc~timali-cI‘s’siIIi oIdjfl,jI\ li..i’l .~ipi.mi-i.,

smiiiiii:t(imiii~ mm lii’ ii IILl(’ 5 siiiiiiii.iil/iII iii mliii I —~‘‘—~—li,ti~iiliii’.lits iIiltiitiili,tti’ liii’

~‘iIli IIt’(i)(’ill(Iiiit.~l~iiiIii t;mlilc2 t.i1

ii.11i iii, I — lii’’ iil—j’4111;li ‘s(’I.i(iii iiil.Iisii’iI iii ‘Jill. I iiiii’,l ‘t1i,LIiI’S

‘sIiin.t(i(I \‘j~ji ‘sIiililii iiiii’’li.iiiil’’ miii Ii, ciii’tfiuiyiils ‘ti—i— ~ mm i—,tt,——id—cli,tui—’i f,,riii imIIu-m tutu first—

iIIIP-u-u,ci- fmuiuum 2 iIi• tlini~tii,I~ ‘.LiI,tIil( 5 ii~

in 1,-al rather thaim ummumiminul ~iflI’., ud .:~‘s~—Ii—m—m—ieh-’s,tuu in luIO(hk ii

11i_mtj(ifl., h.t’si’ lviii (vu v-vIed

ti,ur,lz,ljimim mr,,I,li-:imsTable 1

The Model

4 4 . RI(1)Vi Cl- ~CM(M.j- ~CE(E. I-il. aije

0 i - In-Sample Estimation: I/i 955—IV/i981

(2, P. C2 - ~ CPE H~E. - ~ C0m* XF: I (absolute value of t-statistic in-o parentheses)

• cPA PA,) ~DUM1 LOUMli4 4

- GDUM2 IDUM2Y 2. Il) t~ 281 . 113 ~ M. 001 ~

20 3m 1691 . 0 I0.06,m i 0t3; AL

1~ cPRL IR - ma n 040 SE 372 OW 213

i04. 5

2 . 121R 112-006 ~ PE. -008 ~ IX. XF:l14 AS. ~ ~PERS (PB .i - GMRS M,~ ~327i i 1 (500) i - 0

- 113 PA, 080 DUM1. ‘ 172 DUM2,16 1 ~4L (1 33~ t2 7gm

- ~ ~XRSmX. .1i P 076 SE ‘.28 DW 200 p 016

16 20- ~ cPRS li’ ,m - ~4. l3~RI., 087 ~ P.

0 m350j i 0

I5)U. UF. GGIGAR . cGIcGAR I A 012 SE 032 DW 176 ~ 100

21 2 16Ibm PA, ~ cPRL m~. i Mi RS. 005 ~ PE, , 008 M, - 077 ~ X,.

ii ‘ i m284

j m3221i 0

mfl Y,1P. 100; IX,, 16

-097 ~P,.W; V. itY.V..i’ m 00 1562m i 0

9). flX.X 1100 P - 032 SE 090 OW 83 089

lam F~. ~P. P 1)100 ~ U. UF. O29GAP, O14GAP.

GAP. IIXF. Xi XF,m 100 114 85J In 84~

V2~XR IIXF,X. 11100 pr 070 SE 0’9 ,~- 133 ~. 044

V nominal GNPM money Stock 1Mm;E h’gh employment expendituresP GNP deflator l1972 10cm;PE re!ativc price o~energyX output in 972 domlarsXE - potential oulpjt 1Ra~cbeiatorniAL - corporate bone rateIRS commercia: papcr raleU umiemoloyment roleUF unemployment rate at fum. employment

24